Image Denoising Based On Residual Learning And Convolutional Network And Its Applicatio

Images are very easy to be blurred by the interference of noise in the process of generation and preservation,and low-quality images will increase the difficulty of information acquisition,so image denoising technology has become a hot research area in image processing.At present,the main solutions for image denoising can be divided into traditional denoising algorithms and deep learningbased denoising algorithms.The traditional denoising algorithms are less computationally intensive and faster than the latter,while the deep learning denoising algorithms are more complex and have certain hardware requirements,but they are often better in terms of image texture information retention and noise removal ability.This paper analyzes the existing deep learning denoising algorithms,studies the denoising effect of convolution neural network in shallow and deep convolution layers,and applies the residual learning model to image denoising,proposes a parallel network denoising algorithm based on residual connection,improves the denoising ability and speeds up its speed,and applies the new method to the denoising of water gauge water level image to realize the automatic acquisition of water gauge water level information,with obvious effect.The main work is as follows:1.The denoising effect of convolution neural network in shallow and deep convolution layers is studied.It is found that the number of features extracted from the image is relatively small for the network with shallow layers,and the denoising effect is significantly improved for the network with deep layers.However,the simple deepening of the number of network layers does not significantly improve the denoising effect,and the problem of network degradation cannot be solved.Then the residual learning model is applied to image denoising.Compared with the model that only uses CNN network for denoising,the PSNR value of the denoised image based on residual learning is higher,and optimize the denoising performance of CNN network.2.Aiming at remedy the defeat that most deep convolutional neural networks have the problem of deep level and deep degradation of performance,a parallel network method of feedforward denoising convolutional neural network combined with batch renormalization denoising network is proposed.The new method is to separate the primitive network layer into two parts,and obtain more features by increasing the width of the network rather than the depth Deep structure.Then the residual learning and batch normalization methods are used to improve the denoising quality and accelerate the training.The results show that the new method has more prominent denoising effect and greatly reduces the denoising time.3.In order to better verify the application of parallel network denoising algorithm based on residual connection in practice,the new method is applied to the detection of water level line,denoise the water level image of water gauge,and realize the automatic acquisition of water level information of water gauge.In order to reflect the denoising effect in different environments,this paper gives three different actual environment water gauge water level images.When the noise level is 25 and 50 respectively,the denoising performance is compared with BM3 D algorithm,convolution neural network denoising algorithm and residual network denoising algorithm.The results show that the denoising effect of the new method is better,and the noise of water gauge water level image can be removed,It can also effectively maintain the detailed information of the water gauge water level.